Farouk Miled

Building a Digital Twin Dashboard for actionable infrastructure risk decisions from satellite imagery analysis and simulation of hazard impacts. Developing an agentic system for decision-making assistance based on dashboard insights.

Partnered with DPHi Space to ship a containerized, sandboxed solution for on-board satellite position calibration. Supervised a satellite collision-avoidance project using reinforcement learning.

Designed and shipped an end-to-end hand-object interaction detection model for egocentric video, outperforming the CareEgo baseline by ~8% F1. Owned the ML lifecycle end-to-end and built production-grade RAG and agentic systems.

Built an end-to-end ML workflow predicting airline payment acceptance, lifting acceptance rate by 0.3 points in a pilot. Processed large-scale transaction records using PySpark and Databricks and produced explainable-AI analyses.

Designed a custom latent diffusion model to super-resolve high-dimensional CT scan volumes using MONAI. Delivered higher-fidelity 3D reconstructions for downstream physics simulations with TotalEnergies.

Deployed end-to-end sports video analysis pipelines for French Table Tennis and Swimming Federations. Built detection and tracking models (YOLO, MaskR-CNN, U-Net) and reconstructed 3D racket models via SIFT.

M.Sc. in Machine Learning with core coursework in advanced machine learning, reinforcement learning, deep learning, and multi-agent systems.

General Engineering Degree with coursework in mechanics, fluid mechanics, quantum physics, computer science, applied mathematics, and machine learning.